Model details
gpt-oss-120b
The gpt-oss-120b model is built upon a scalable transformer architecture specifically engineered to handle complex, multi-step reasoning tasks and agentic workflows. Designed as a general-purpose solution for developers, it balances high-stakes performance with practical deployment needs. The model features a unique capability for configurable reasoning effort, allowing users to adjust performance based on specific latency requirements. By providing full access to its chain-of-thought process, it offers increased transparency for debugging and verifying outputs, making it a robust choice for building durable, production-ready systems that require reliable instruction following and structured data generation.
Training for this model involved a combination of reinforcement learning and techniques derived from advanced frontier systems, ensuring it achieves performance levels comparable to proprietary benchmarks. It utilizes a specialized harmony response format, which is essential for its operation and integration into agentic pipelines. Beyond its core reasoning strengths, the model has undergone extensive adversarial fine-tuning to improve safety and reliability in specialized domains such as cybersecurity and technical troubleshooting. With its ability to run efficiently on standard high-end hardware, it provides a flexible, permissive foundation for developers to customize and deploy sophisticated AI applications without the constraints of closed-source environments.
Quick Info
Powered by- Provider
- OVHcloud AI Endpoints
- Model key
- gpt-oss-120b
- Release date
- Aug 28, 2025
- Last updated
- Aug 28, 2025
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.09
- Output token cost
- $0.47
Limits
- Output tokens
- 131,072 tokens
- Context window
- 131,072 tokens